The default project for the course is a 5DOF robotic arm, with a second project being a mobile robot (Turtlebot). The robotic arm project, as the default project, is already incorporated into the homeworks. The mobile robot project is not, but has a mostly established program of study.

1. Salamander
Build and program a salamander robot based on the Biorobotics Laboratory design called Pleurobot, some of whose details were recently published. Involves mechanical design and construction, kinematic modeling of articulated robots, and servomotor programming.

2. Mechbot
Re-design and program a mech-warrior type robot from a previous iteration of the course. The leg design is more like that of a bird, which has an inverted knee (relative to humans). Involves mechanical design and construction, kinematic modeling of articulated robots, and servomotor programming. Also may involve optimal trajectory synthesis methods for movement. Initially, planar models would be explored, followed by spatial models.

3. Biped/Humanoid Robot
There is a small bipedal humanoid robot, Bioloid-GP with some custom modifications, available for exploring bipedal movement and locomotion. This project would be about learning to model and control biped robots. The initial exploration would start with a planar model of the biped and ramp up to a full spatial model. The idea being to understand the kinematics of balance and of walking for biped robots. Please let me know if this is of interest, as there is one student already committed.

4. Factory
Create a factory lab for the class. The class originally had a small industrial workfloor courtesy of a, now retired, professor who was using his research laboratory as the class laboratory. It was non-functional when I first arrived, and a replacement has never been created. Let's see if we can create our own miniature factor floor complete with a machine vision overhead view for planning and control. This project will involve creating miniature conveyor belts, revising an embedded computing circuit design and programming the microcontroller, and software design. Creating the mini-factory will involve using your knowledge of kinematics in the design process.

5. NASA Space Challenge
NASA has this space robotics challenge that involves getting a humanoid robot to perform certain tasks. The humanoid robot is NASA's Valkryie robot. The main website has more details. I have no idea what this entails since one must first pre-register in order to get details. What I imagine is that they will do something like DARPA's Robotics Challenge. You will be given access to a robot simulator (to be downloaded to your computer, and in all likelihood requiring linux) which will be used to program the robot to perform different tasks. If you are lucky and it builds on ROS, then programming might be possible in python. There will most likely be a mix of tele-operation and autonomous operation. Outside of this, I really don't know. I presume then, this is for the motivated team.

6. Johnny 5 like robot: AWSM The RICAL Lab here at Georgia Tech has this robot whose design is similar to Johnny 5, from the 80's movie of same title. The robot works but is tele-operated and has mostly manual operation. It would be cool to start to give it some smarts. (NOTE: This project is a maybe at the moment. I need to confirm with Dr. Cho that something reasonable can be done. I should hear back by Monday or Tuesday (September 5/6). To see AWSM the Robot, go to the RICAL page and flip the image slideshow to the right a few times. The 6th or 7th image gives a good full robot view of AWSM.

7. Optimal control: IPOPT + python or IPOPT + Matlab
The latest trend in robots and control of them is to utilize optimal control strategies to generate control signals for the robots. We are interested in enhancing and porting an existing optimal trajectory generation package Optragen. Porting would involve translating the current design to python and incorporating IPOPT or some other similar nonlinear optimization solver. A slightly simpler step would be to enhance the current version of Optragen to work with IPOPT's Matlab version (or even with Matlab's internal optimization solvers). In the process, more complex optimization problems will be solved than the current examples. If possible, they can even be implemented on actual experimental robot platforms (presuming that it is fairly simple to do so). As a final consequence, Optragen will be able to run on more platforms and on more optimal control problems than it can now.

Students can define their own project, through discussions with instructor. Usually the discussions are to assess the appropriateness of the chosen topic, both in terms of its relation to the course material and in terms of its level of difficulty.

Spring 2015

Main